Since 2014, integrating predictive maintenance and digital twin technologies has attracted many industries. The predictive maintenance method based on digital twins (PdMDT) heavily relies on real-time communication between the twins. Compared with traditional predictive maintenance, the key difference of PdMDT is that it can real-time perception, regulation, and prediction. Traditional predictive maintenance modes based on data analysis can realize the qualitative analysis of monitoring data. There is still a large array of equipment online status, environmental parameters, and historical records left out from the analysis, which affects the accuracy of the analysis.
Some of the most difficult challenges for industrial companies are scheduling complex manufacturing lines, maximizing throughput while minimizing changeover costs, and ensuring on-time delivery of products to customers. Digital twins and AI can help through their ability to consider a multitude of variables at once to identify the optimal solution. For example, in one metal manufacturing plant, an AI schedule agent reduced yield losses by 20 to 40 percent while significantly improving on-time delivery for customers. Companies must establish an environment where AI scheduling agents can learn to make good predictions to achieve this goal. Relying on historical data and machine learning is simply inadequate since the agents cannot anticipate future issues. Instead, organizations can start by building a simulation or “digital twin” of the manufacturing line and order book. A scheduling agent can then schedule the line. The agent’s performance is scored based on the cost, throughput, and on-time delivery of products.
In essence, the perception ability of PdMDT is the outcome of relevant maintenance data acquisition processes. Embedded network devices (such as sensors) and interconnected communication devices (such as the Internet of Things and wire-and-wireless networks) facilitate collecting the operation status, parameters, and environmental conditions relevant to the target machine. In this regard, the data collected from various sources enters pre-processing operations, including noise reduction, segmentation, feature extraction, and selection. Then, this real-time pre-processed data is retrieved and compared to the machine’s fault knowledge base, including the machine’s historical maintenance information.
Considering digital twins, there are different levels of connectivity between the physical world and its digital counterpart. A “shadow” has automated one-way data flow between the physical object and its digit clone. Change in the state of the physical object leads to change in the state of the digital object, but not vice versa. Thus, a shadow receives real-world data automatically yet does not actuate changes back into physical reality autonomously.
The digital shadow is, therefore, easier to implement, and it is by far the best option to start on such a digital transformational journey. As the intelligent algorithms powering the digital clones require a training phase to learn the characteristics and behaviors of their physical equivalents, and this type of learning, in turn, is equally often supported via human interaction, shadows are the most risk-averse entry point into digital twining at its finest.
However, the market’s preferred misuse of digital twin terminology – the model, or simulation, is a digital representation of an existing or planned physical object that does not use any form of automated data exchange or integration between the physical object and its digital cone. A change in the state of the physical object has no direct effect on the state of the digital, and vice versa, and thus, the model would need to be manually updated regularly to avoid discrepancies.
Asst. Prof. Suwan Juntiwasarakij, Ph.D., Senior Editor & MEGA Tech